The key word in “Data Science” is not Data, it is Science
simplystatistics.org
simplystatistics.org
My modest exposure to machine learning at the professional level gives me impression that the "real experts" combine a strong mathematical understanding, long experience and some good rules of thumb to perform better than a grad student shooting in the dark, if they happen to perform better.
Oddly enough, all articles about how hard it is to become a "real data scientist" gives the impression that however much expertise is involved, that expertise isn't the codified understanding that is "real science" - even a physics undergraduate does real physics because scientists, physics codified their methods.
Maybe "data science" can become science. But suspect that what will become scientific is the understand of whatever entity is producing the data. Which isn't to discount the learning of experts here but simply to note that compendiums of rules-of-thumb and feelings indicate what Thomas Kuhn might a pre-scientific field.
http://en.wikipedia.org/wiki/Least_squares#History
In one sense science is just the outcome of correctly applying the scientific method to a given problem, and in that sense data science is perhaps a meta-science, in that it focuses on the part of the scientific method concerned with making inferences from data, and tries to improve it's effectiveness, using process that is itself scientific to some degree (applying mathematical reasoning to find more reliable ways of inference, introducing additional empirical experiments like cross-validation into the inference process itself for confirmation). I think comparing physics to statistics is a categorical error, and at the same time it's worth keeping in mind that modern physics, biology and chemistry certainly rely very heavily on probability theory and mathematical statistics, so those disciplines always brought up as "properly scientific" are in many respects only rigorous to the extent statistics is.
I think the data component is a feature of applied physics in general. Physics problems have had a tendency to be extremely data driven, because when you're dealing with things out in the cosmos, or trying to model nuclear structure for the first time, or shed light on the Higgs mechanism, the data is all you have. You often can't see it with your eye, or a microscope. The effect you're observing only lasts a femtosecond, so you conceive of extremely fancy contraptions that measure the secondary effects en masse (if you'll allow the pun), and you end up with fantastically large data sets.
That said, when I look at the scientists I knew/know, including astronomers, their solutions often take a different approach that what I see today in the "data science" blogosphere. There, the approach seems highly centered on scaling various methods to work with huge amounts of data, whereas most of scientists seem to do just the opposite. They go to great lengths--such as building massive integrating antennae the size of football fields--to make sure they collect precisely the correct data, and nothing more. They then take that data and apply the technique that they know must work, if their hypothesis is correct.
Web 2.0 was a trend in prominent personality types, types of websites, business models, increased scale of online interaction, use of real names, web design styles, programming languages. It wouldn't have been absurd to see a guy with a specific look and a specific laptop and say "that's so web 2.0". To add insult to injury, pretty much all the stuff that gets captured in a word like web 2.0 has predecessors. Crowd sourcing? What about Wikipedia?
I think a good analogy is "movement" in art, philosophy & culture. Modernist is a word that encompasses Frank Lloyd Write, Pablo Picasso, James Joyce, Ayne Rand & karl Marx. It applies to paintings, manifestos, econometrics and buildings with straight lines.
That's the kind of word that 'data science' is. We found ourselves recording a lot of data as a sort of side effect of digitization. It's growing. Then we start to try and get some value from that data. Some new stuff is possible with that volume of data. Some new people are now interested in data. A lot of the tools people were using to collect and analyze data don't work at that volume, so we start using new tools. We end up with a word that includes astronomers, netflix, medical researchers, self driving cars, R, statistical theories etc.
Data science doesn't mean anything that specific yet. It's best not to lead the discussion (as I am doing right now) to a discussion about the word, what qualifies as data science.
1. There is science, and there is data. Science has data, and is a thing you can do. Data is just data. "Data Science" as a description of an activity is like saying "Words Writing" or "Money Banking". As a whole the term holds no concrete meaning, thus it is contorted to whatever the current topic of conversation is (this is exactly what you've just said, I think).
2. There has been a massive change in the volume of data and the tools we use to analyze it. This is nothing to do with what you're doing, what field you're in, or whether you know what you're doing. Again, data is just data. (Also what you said).
3. If you want to use the term in your conversations, blog posts or--if you must--publications, fine. Just make sure YOU have already set in stone what it is you're talking about. It shouldn't take others probing your sentence structure, nomenclature and argumentation for you to come up with a precise definition of what you mean. That's just bad SCIENCE, no matter how much DATA you have.
The "cloud" buzzword really pissed people off. I thought that was an overreaction too. If you say that enterprise software companies need to deal with these new cloud competitors, you mean a whole cluster of things. Web software, SaaS pricing models and sales cycles, smaller companies that don't run their own infrastructure, etc.
Good article, nevertheless.
This isn't how it should be, and it isn't universal, but based on my limited sample set, BAs tend to be folks who can write what someone else tells them, but they rarely contribute much more. (The ~10% who defy this description are worth their weight in gold, perhaps platinum)